Taxi detour behavior analysis is a crucial topic in uncovering and interpreting the complicated interaction between diverse traffic context and dynamic routing efficiency. In this paper, a Map-matching based detour classification framework is proposed. Supported by a well-established multi-layer road index (MRI) system in Shenzhen, China, we processed one week of taxi trace trajectories along with both the visualization of related spatio-temporal distribution features of detour behavior at DRS level and some statistics of several key travel indicators. Results reveal that (i) taxi detouring should be a very common routing behavior and 27.5 percent of taxi trips made a detour larger than 1.5 kilometers compared with their corresponding shortest path distance; (ii) both the level of detour intensity and ratio are highly affected by road features and dynamics in different spatio-temporal interaction patterns.

CICTP